基于单目视觉的同时定位与地图构建方法综述  

A Survey of Monocular Simultaneous Localization and Mapping

作　　者：刘浩敏[1] 章国锋[1,2] 鲍虎军[1] 

机构地区：[1]浙江大学CAD&CG国家重点实验室,杭州310058 [2]浙江大学工业信息物理融合系统协同创新中心,杭州310058

出　　处：《计算机辅助设计与图形学学报》2016年第6期855-868,共14页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家科技支撑计划(2012BAH35B02);国家自然科学基金(61232011;61272048);中央高校基本科研业务费专项资金(2015XZZX005-05);全国优秀博士学位论文作者专项资金资助项目(201245)

摘　　要：增强现实是一种在现实场景中无缝地融入虚拟物体或信息的技术,能够比传统的文字、图像和视频等方式更高效、直观地呈现信息,有着非常广泛的应用.同时定位与地图构建作为增强现实的关键基础技术,可以用来在未知环境中定位自身方位并同时构建环境三维地图,从而保证叠加的虚拟物体与现实场景在几何上的一致性.文中首先简述基于视觉的同时定位与地图构建的基本原理;然后介绍几个代表性的基于单目视觉的同时定位与地图构建方法并做深入分析和比较;最后讨论近年来研究热点和发展趋势,并做总结和展望.Augmented reality (AR) is a technique that allows to seamlessly composite virtual objects or informa-tion into real scene. Compared to traditional text, images and videos, AR is a more effective and intuitive way for information presentation and has wide applications. Simultaneous localization and mapping (SLAM) is a key fundamental technique for augmented reality, which provides the ability of self-localization in an unknown envi-ronment and mapping the 3D environment simultaneously. The localization and mapping enables fusion of virtual objects and real scenes in a geometrically consistent way. In this paper, we describe the basic principles of Visual SLAM, and introduce some state-of-the-art monocular SLAM methods with deep analysis and comparison. Fi-nally, we discuss some research tendency in recent years and make conclusions.

关 键 词：增强现实 同时定位与地图构建 运动推断结构 多视图几何 摄像机跟踪 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]